Dekkera bruxellensis CBS2499 was used in this study. Its whole genome sequence is available at http://genome.jgi.doe.gov/Dekbr2/Dekbr2.home.html (Piškur et al., 2012). Cells were stored in YPD medium (10 g/L yeast extract, 20 g/L peptone, 20 g/L glucose, 5.5 pH) supplemented with 20% (v/v) glycerol at -80°C. Cell revitalization was performed inoculating the glycerol stock at 1% (v/v) in YPD broth. Cultures were placed into an incubator (Heidolph, Schwabach, Germany) at 30°C for 3 days.

Experiments were run to collect yeast biomass for RNA extraction, retrotranscription and the analysis of gene expression by real-time quantitative PCR (qPCR). All fermentations were carried out in simil-wine Medium (SWM) [2.50 g/L glucose, 2.50 g/L fructose, 5 g/L glycerol, 5 g/L tartaric acid, 0.50 g/L malic acid, 0.20 g/L citric acid, 4 g/L L-lactic acid, 1.70 g/L yeast nitrogen base w/o AA and ammonium sulfate (Difco, Sparks, MD, United States), 0.005 g/L oleic acid, 0.50 mL tween 80, 0.015 g/L ergosterol, 0.020 g/L uracil, 0.010 g/L p-coumaric, 0.010 g/L ferulic acid, and 1.50 g/L ammonium sulfate]. Variants of SWM were prepared at different molecular SO₂ (below: SO₂) and ethanol concentration and pH value, adjusted with NaOH, depending on the conditions set by the chosen RSM (Table 1). Media was sterilized with 0.20 μm cellulose-nitrate filters. Cultural media were stored at 22°C prior the cell inoculation. SO₂ was added immediately before the inoculum from a 4 g/L sodium metabisulphite in mQ water. The theoretical content of molecular SO₂ was calculated according to Usseglio-Tomasset and Bosia (1984), Ribéreau-Gayon et al. (2006), and Duckitt (2012). Cellular growth was monitored by OD at 600 nm. Fresh cells in YPD broth were centrifuged at 3500 rpm for 15 min (Hettich, ROTINA 380R, Tuttlingen, Germany); then, cells were washed in 0.9% (w/v) NaCl and inoculated at 0.1 OD_{600 nm} in flask in SWM adjusted at 5% (v/v) ethanol, pH 4.5 and maintaining an air/medium ratio of at least 40% in order to ensure aerobic condition. Cellular pre-cultures were grown at 25°C for 3 days, in aerobic condition. An aliquot of the fresh cultures was analyzed by plate count to calculate the exact number of viable cells transferred into each variant of the SWM for the RSM (Table 1). The inoculum was carried out at 0.25 OD_{600 nm} in SWM modified as required by the RSM scheme (Table 1). The inoculated media were divided into 10 mL aliquots in sterile and hermetically closed tubes with no headspace volume, and cultivated at 22°C in static condition. Each aliquot sample was used once for analyses. Cellular growth was monitored daily by total plate count and OD_600nm measurement. At 1.00 ± 0.2 OD_{600 nm} cells two aliquots were pelleted by centrifugation (11000 rpm, 1 min, 4°C) (Hettich, ROTINA 380R, Tuttlingen, Germany), collecting a total cell amount of 20 OD_{600 nm}, immediately frozen with liquid nitrogen and stored at -80°C until use. For the RSM scheme, the cultures were arranged according to the chosen experimental design.

The extraction of total RNA from pellets was carried out using Presto Mini RNA Yeast Kit (Geneaid, New Taipei City, Taiwan) with few modifications. Briefly, cell lysis through mechanic disruption was performed in 500 μL Buffer RB, 5 μL β-mercaptoethanol, and an iso-volume of glass beads (425–600 μm, 154 Sigma–Aldrich, Saint Louis, MO, United States). Three breaking cycles with TissueLyser (Qiagen, Hilden, Germany) for 2 min at the maximum oscillation frequency, interchanged with 1 min on ice, were applied. The supernatant was centrifuged at 16000 × g for 3 min (Hettich, Tuttlingen, Germany). The genomic DNA residue was degraded using 100 μL of 2 KU/mL DNase (Sigma–Aldrich, St. Louis, MO, United States) for 15 min at room temperature. Following steps were carried out according to the manufacturing’s instructions. RNA concentration was determined by measuring the absorbance at 260 nm (BioTek, Winooski, VT, United States). The integrity of RNA sample (0.3 μg RNA, 2 μL RNA loading Buffer 5X, H₂O DEPC up to 10 μL) was assessed, after 5 min treatment at 65°C, by electrophoresis on 1.2% agarose gel [90 mL DEPC water, 10 mL 10X formaldehyde gel buffer (200 mM MOPS, 50 mM sodium acetate, and 10 mM EDTA)] adjusted at 7 pH with NaOH prepared in DEPC water 37% (v/v) formaldehyde added. The electrophoretic run was carried out at 100 V for 1 h and bands were UV visualized (Bio-Rad, Berkeley, CA, United States). RNAs were stored at -80°C until cDNA synthesis. The RNA retrotranscription was obtained with the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). cDNAs were stored at -20°C until used for the qPCR assays.

Five genes, pyruvate decarboxylase (PDC) (DbPDC), aldehyde dehydrogenase (DbALD), actin (DbACT), eukaryotic translational elongation factor (EF) (DbEF), and tubulin (DbTUB), were analyzed to identify a HKG suitable in the normalization process of the gene expression of CD (DbCD) and VPR (DbVPR) (Table 2). Gene sequences of DbPDC and DbALD were identified using S. cerevisiae S288C (Schifferdecker et al., 2014), Komagataella phaffii CBS7435 and GS115, D. bruxellensis CBS2499 (Piškur et al., 2012), and B. bruxellensis AWRI1499 (Curtin C.D. et al., 2012) genomes. SGD¹, NCBI², and ENA³ databases were used as sequence sources. All alignments were performed through BLAST and ClustalIX2. Primer pairs were obtained at NCBI website⁴ and validated for no forming neither self nor cross-dimers⁵ (Table 2). The DbCD gene sequence for primer design was deduced by Godoy et al. (2014).

Two sets of gene expression analysis were set up under different oenological conditions: (i) to identify a suitable HKG for gene expression normalization; (ii) to analyze the relative expression of DbCD and DbVPR, by using the gene identified in (i). As far the primers couples designed in this study for DbCD, DbALD, and DbPDC genes, they were also validated by a standard PCR amplification in a 25 μL reaction composed by: 1 U Taq, 200 μM dNTPs (Biotech rabbit, Dusseldorf, Germany), 1X Taq Buffer (Genscript, Piscataway, NJ, United States), 1 mM MgCl₂ (5Prime, Hilden, Germany), 0.1 μM primer forward and 0.1 μM primer reverse (Eurofins Genomics, Ebersberg, Germany), and 80–100 ng DNA. The amplification cycle was: 95°C for 6 min, 95°C for 45 s/54°C for 30 s/72°C for 1 min (repeated 34 times), and 72°C for 10 min. Results were visualized on a 2% agarose gel prepared in TAE 1X buffer (20 mL TAE 50X, 980 mL demineralized water) and 0.5 μg/mL ethidium bromide. Electrophoresis was set at 80 V for 1.30 h. PCR products were sequenced by an external provider (Eurofins genomics, Ebersberg, Germany).

As far qPCRs, they were performed in a Realplex Mastercycler EP Gradient Thermocycler (Eppendorf, Hamburg, Germany) using a 15 μL reaction mix composed as follow: 2X SYBR Green Master-Mix (Biotech rabbit, Dusseldorf, Germany), 200 nM-100 nM-50 nM primer forward and primer reverse (Eurofins genomics, Ebersberg, Germany), and 10-fold dilution cDNA. The qPCR amplification cycle was set at 95°C for 30 s, 54°C for 30 s, and 65°C for 30 s; repeated for 40 times. At the end of the reaction (95°C for 15 s), a melt-curve was generated by increasing the temperature from 60 to 95°C, with a step at 0.5°C. All cDNAs were run as technical duplicates in a 96-well plate (Eppendorf, Hamburg, Germany). For each gene, three decimal serial dilutions at least were prepared into DNA LoBind tubes (Eppendorf, Hamburg, Germany) and stored at -20°C. The amplification curves were analyzed with Realplex software (Eppendorf, Hamburg, Germany).

The 2^-ΔΔC_T method was applied on the basis of Livak and Schmittgen (2001) to calculate the relative expression of DbCD and DbVPR respect the chosen HKG expression. Results were expressed as fold-changes whereas the expression value of the target gene (normalized against DbTUB expression) was expressed as increase or decrease respect to its expression in the calibrator (for equivalent amount of samples) corresponding to the growth condition “LS” [0 mg/L mol. SO₂, pH 4.5 and 5% (v/v) ethanol] described in the paragraph “Gene Expression Stability.”

The expression of DbPDC, DbALD, DbACT, DbEF, and DbTUB genes was evaluated setting up a qPCR multiplex assay under two different oenological conditions of the SWM called “low-” and “high-” stringent (LS and HS, respectively) growth conditions. In particular, the LS condition was characterize by 0 mg/L mol. SO₂, pH 4.5 and 5% (v/v) ethanol while the HS condition by 0.25 mg/L mol. SO₂, pH 3.5 and 12.5% (v/v) ethanol. Yeast cultures were prepared in duplicate; three RNA extractions and the following cDNA synthesis were performed from each independent culture.

GeNorm analysis (Vandesompele et al., 2002) (Genex software version 4.3.6, MultiD analyses, Gothenburg, Sweden) was used to determine the stability of gene expression (termed M-value), by analyzing each reference gene against the others in a pairwise variation that serially excludes the least stable gene (highest M-value) from the analysis. At the end, genes are ranked with an accepted cut-off value of 0.50 according to their expression stability. Normfinder algorithm (Genex software version 4.3.6, MultiD analyses, Gothenburg, Sweden) separates the variation into an intra-group and an inter-group contribution. The analysis is repeated without considering the groups and this, estimates a robust standard deviation (SD) for each gene. The accumulated standard deviation (Acc. SD) is a reliable indicator of the number of reference genes to be used. All the genes were analyzed in the same assay to reduce any further experimental variability.

In order to investigate the expression of DbCD and DbVPR genes and the production of VPs in oenological conditions a Box–Behnken experimental design and RSM were applied. SWM samples were formulated with different level % ethanol (v/v) (5 – 8.75 – 12.5), pH values (3.5 – 4.0 – 4.5), and molecular SO₂ (mg/L) (0 – 0.125 – 0.25) (Table 1). The 15 trials provided by Box–Behnken experimental design were analyzed using Statgraphics Plus 5.1 software. The expression values of investigated genes were normalized with the HKG expression.

The fit of the model was evaluated by the linearity coefficient (R-squared). The regression approach was used to determine the effects produced by SO₂, pH, and ethanol variables. The main effects (A, B, and C) and both the linear (AB, AC, and BC) and quadratic effects (AA, BB, and CC) were statistically validated by analysis of variance. To identify the most important factors, a standardized Pareto chart is drawn. In particular, each effect is converted to a t-statistic by dividing it by its standard error (data not shown). These standardized effects are then plotted in decreasing order of absolute magnitude. Statistically relevant effects with a p-value less than 0.05 (95% confidence level) were reported in a response surface graph where the three-dimensional surface is described by a second-order polynomial equation.

The content of hydroxycinnamic acids, namely p-coumaric and ferulic acids, vinyl phenol, vinyl guaiacol, ethyl phenol, and ethyl guaiacol in the cultures of the 15 runs of Box–Behnken experimental design was assessed in the obtained samples by an Acquity HClass UPLC (Waters, Milford, MA, United States) system equipped with a photo diode array detector 2996 (Waters). Chromatographic separations were performed with a Kinetex C18 150 mm × 3 mm, 2.6 μm particle size, 100 Å pore size (Phenomenex, Torrance, CA, United States). Eluting solvents were (A) trifluoroacetic acid 0.05% (v/v) and (B) methanol. The gradient program was 0.1 min, 20% B; 0.1–2 min, 35% B; 2–14 min, 58.5% B. The separation run was followed by 7 min of column rinsing and conditioning. The flow rate was 0.5 mL/min and the column temperature was 28°C. The samples were filtered with PVDF 0.22 μm filter prior the injection. Calibration curves were obtained for p-coumaric and ferulic acids, vinyl phenol, vinyl guaiacol, ethyl phenol, and ethyl guaiacol concentrations in the range from 0.1 to 20 mg/L. Quantification was performed according to the external standard method. Data acquisition and processing were carried out by Empower 2 software (Waters) at 320, 280, and 260 nm for hydroxycinnamic acids, ethyl phenols, and vinyl phenols, respectively. Yield values of VPs were calculated as the molar ratio between each product (vinyl phenol, vinyl guaiacol, ethyl phenol, and ethyl guaiacol) and the corresponding hydroxycinnamic acid potentially used as substrate. Data were analyzed with Statgraphics Plus 5.1 using the RSM approach.

DbPDC gene was identified in the scaffold 1 at 1700 bps (e_gw1.1.1485.1) of D. bruxellensis CBS2499 genome; in particular, the nucleotide sequence showed about 55% identity with the S. cerevisiae genes encoding for PDC1, PDC5, and PDC6 (55.1, 55.8, and 55.5%, respectively). Due to the similar level of identity found among the three isoforms, PDC1 sequence was chosen for a further investigation in the genome of B. bruxellensis AWRI1499. The nucleotide sequence with accession number “EIF49850.1” was identified as a possible homologous of S. cerevisiae PDC gene with an identity of 55% (identity of 96.9% with e_gw1.1.1485.1). In K. phaffii genome, the gene codifying for KpPDC showed two potential isoforms differently located in K. phaffii CBS7435 (chromosomes 3 and 4). Only the sequence on the chromosome 3 identified the homologous gene (identity of 100%) on the genome of the strain K. phaffii GS115, with accession number XM_002492352.1. Thus, this gene was aligned against D. bruxellensis CBS2499 and the sequence in the scaffold 1 (e_gw1.1.1485.1) was confirmed as the potential homologous gene of KpPDC (55.5%identity). In conclusion, the open reading frames represented by the accessions e_gw1.1.1485.1 and EIF49850.1 of D. bruxellensis CBS2499 and B. bruxellensis AWRI1499, respectively, were identified as the homologous genes of ScPDC1 and KpPDC.

As regards DbALD, among the three genes (ScALD3, ScALD2, and ScALD6) encoding for the sequence of ScALD6 of S. cerevisiae S288c genome led to the identification of a possible homologous gene in D. bruxellensis CBS2499 genome in the scaffold 4 at 1523 bps (e_gw1.4.403.1) with an identity of 55.6%. ScALD6 sequence was also aligned against the genome of B. bruxellensis AWRI1499 and the resulting amino acid sequence with the accession number “EIF46557.1” showed an identity of 56% (99.4% identity with e_gw1.4.403.1). In K. phaffii genome, the gene encoding for KpALD was identified on different chromosomes; the nucleotide sequence in the scaffold 20 at 1496 bps (e_gw1.20.29.1) of the chromosome 3 of the strain CBS7435 showed the highest identity (67.8%) with both e_gw1.4.403.1 and EIF46557.1 open reading frames of D. bruxellensis CBS2499 and B. bruxellensis AWRI1499 genome, respectively. Thus, these last genes were used for primer design being considered the homologous genes of ScALD6 and KpALD.

The primer pairs designed on DbALD, DbPDC, and DbCD were evaluated for their ability to produce a specific fragment through a standard PCR and further sequencing of the amplified products. A unique amplification product of 140 bps for all the three genes investigated was obtained (data not shown). This value corresponds to the expected product length on the base of the size (Table 2) of in vitro primers design. No aspecific products were detected and no amplification was observed with S. cerevisiae S288C and K. phaffii GS115 used as negative controls. Primer specificity was confirmed by sequencing with a 100% identity with the target sequences.

All primers designed for the amplification of the potential HKGs (DbALD, DbPDC, DbEF, DbTUB, and DbACT) and the target genes (DbCD and DbVPR) were validated to assess whether the qPCR reactions were really optimized. Five dilutions of cDNA samples obtained from cell culture of D. bruxellensis CBS2499 grown in SWM at LS condition [0 mg/L SO₂, pH 4.5, 5% (v/v) ethanol] were tested to evaluate the ones containing from 10³ to 10⁶ copies of template that were able to give amplification curves between 30 and 20 C_T values, respectively. The obtained C_Ts values were relatively low and similar; the lowest one (about 13) was given by DbEF gene, while the highest (about 20) was obtained for DbCD gene, thus revealing similar expression levels among the amplified genes. Then, a standard curve was created to assess primer efficiency of both the target genes and potential HKGs, as well as to be used as “standard” within the normalization plate used for HKG identification by qPCR. The R² values obtained for all primer pairs ranged from 0.980 to 0.999.

Five genes were evaluated for this purpose (Table 2): two genes encoding for metabolic enzymes, PDC and acetaldehyde dehydrogenase (ALD), were chosen based on their important role on fermentative metabolism and on NAD(P)H supply. The three others, encoding for EF, tubulin (TUB), and actin (ACT), have been already used as HKG in other studies (Nardi et al., 2010; Rozpędowska et al., 2011; Moktaduzzaman et al., 2016).

DbALD, DbPDC, DbEF, DbTUB, and DbACT were analyzed by a qPCR multiplex assay to identify the reference gene with a constant expression level across the experimental conditions under study. Expression stability of potential HKG genes were assessed at the two extreme growth conditions of the used experimental design, LS [0 mg/L SO₂, pH 4.5, 5% (v/v) ethanol] and HS [0.25 mg/L SO₂, pH 3.5, 12.5% (v/v) ethanol]. The cultures showed a negligible lag phase reaching a similar final biomass (1.4–1.7 OD_{600 nm}) in 8 days. The absolute quantification approach was employed to obtain the qPCR results from the assayed normalization plate. Thus, a direct comparison between C_Ts of each sample and C_Ts of the standards (corresponding to the transcript copy number of each serial dilution of the HKG candidates) was accomplished. Overall, genes presented C_Ts spanning from 11 to 20, with DbEF and DbPDC having the lower values (Table 3). C_T data were submitted to GeNorm (Vandesompele et al., 2002) and Normfinder analysis. Because of the elimination process, GeNorm algorithm cannot identify an optimum reference gene and ended up by suggesting a pair of genes having the best same M-value of 0.186, DbACT and DbTUB (Table 3). For a single gene discrimination, Normfinder was employed along with GeNorm algorithm. Since samples came from two different treatment groups, Normfinder algorithm separated the variation into an intra-group and an inter-group contribution. The analysis was then repeated without considering the groups and this allowed to estimate a robust SD; the lowest SD (0.0929) was assigned to DbTUB (Table 3). A minimal value of the accumulated standard deviation was a great indicator of the optimal number of reference genes to be used for normalization. The highest expression stability revealed by DbTUB, attributed by both the lowest M-value and the SD, identifying this gene as the HKG for this study.

Real-time qPCR assays were carried out to test all conditions of the experimental design in order to study the role of SO₂, pH, and ethanol on DbCD and DbVPR genes expression. All the assays produced amplification curves in the range of the best sensitivity of the qPCR (20–30 C_T values) and a high reproducibility within a single test and among tests was obtained; indeed, an overlapping of the amplification curves of the replicates of both each run and the calibrator was observed. This was particularly evident in the case of DbTUB amplification that showed a constant gene expression (C_T value of 23) among the 15 conditions evaluated, confirming once again its reliable role as HKG.

Although the experimental design has to be considered functional to only apply the RSM approach and data cannot be individually interpreted as not obtained from biological replicates (except for runs 13, 14, and 15), it was possible to observe that DbCD gene was downregulated in all the tested conditions with fold-change values ranging between 0.14 and 0.66 (Table 1). The application of the Box–Behnken results to the RSM approach allowed to analyze how the DbCD gene expression was influenced by SO₂, pH, and ethanol by predicting further expression values inside the environment of the tested variables. Indeed, as regards the DbCD gene expression, a high R-squared values indicated a good fit of the model to the experimental data explaining the 98.3% (R-squared) of the DbCD gene variability (Table 4). Main and interaction effects (linear and quadratic) of the factors on the gene expressions are reported in Table 5 and shown in the standardized Pareto chart (Figure 1). While pH and ethanol factors produced a significant effect (P-value < 0.01) on the DbCD gene expression, SO₂ did not affect it. On the contrary, linear interactions between SO₂ and pH and SO₂ and ethanol revealed a substantial influence (P-value < 0.05) (Table 5 and Figure 1) thus concurring to define the response represented as three-dimensional surface (Figures 2A,B).

The shape of the surface obtained for SO₂ and pH interaction (Figure 2A) on the response reflected the predominant inhibition by pH, since the expression of the gene decrease rapidly up to pH 4. In particular, the change in DbCD expression occurring from the lowest to the highest level of pH (Figure 3A) was the same for both 0.125 and 0.250 mg/L levels of SO₂; the parallel trend of lines indicated that the effect of the pH on the response is probably not dependent from these SO₂ values. Even when pH was in the range 3.5–4 and SO₂ at 0 mg/L, the observed lines were almost parallel with respect to the other lines (with an overlapping between 0 and 0.250 mg/L of SO₂). On the contrary, when pH was set between 4 and 4.5 a moderate interaction of this factor with SO₂ occurred (lines are not parallel) (Figure 3A).

On the other hand, the interaction between SO₂ and ethanol produced a response that changed faster as function of ethanol (Figure 2B). In detail, considering ethanol from 5 to 12.5% (v/v) and SO₂ at the concentration of 0 mg/L, 0.125 mg/L or 0.250 mg/L, the observed lines were not parallel indicating that an interaction between ethanol and SO₂ exists (Figure 3B). If ethanol at 5% (v/v) interacted with 0.25 mg/L SO₂, the DbCD expression was lower than the one revealed by the condition at 0 mg/L SO₂. This is probably due to the effect of ethanol along with SO₂ in determining more stress to the cell. Moreover, the expression at 0 mg/L SO₂ and 8.75% (v/v) ethanol was slightly lower than the one revealed at 0 mg/L SO₂ and 5% (v/v) ethanol.

The comparison between the two interaction plots (Figures 3A,B) allowed identifying the SO₂-ethanol as the stronger interaction to define the expression of DbCD, as also showed by the p-value of this linear interaction (AC, Table 5 and Figure 1).

Finally, based on the response surfaces for DbCD gene expression and the model equation it is also possible to predict further responses in addition to those obtained in this study; according to this prediction approach, the combination of the factor levels that maximizes the DbCD expression (0.834-fold change) is at 0.25 mg/L, 4.5 and 12.5% (v/v), respectively, for SO₂, pH, and ethanol.

As far the DbVPR gene expression, it showed a different trend in regulation in comparison to DbCD gene. Even if data of the experimental design cannot be singularly interpreted, DbVPR seemed to be upregulated in runs 1, 2, 3, 4, 5, and 10 with fold-change value ranging between 1.11 and 1.80, whereas in the other cases it was slight downregulated, being values lower than 1. Interestingly, following the results in Table 1, although DbCD and DbVPR genes were expressed at their maximum level under the same growth condition corresponding to SO₂ 0.25 mg/L, pH 4.5, ethanol 8.75% (v/v) (run 4). The statistical processing of expression data provided a regression equation of the proposed model with a goodness of fit of 87.3% (R-squared) (Table 4). In this case only a positive quadratic effect of SO₂ and pH resulted statistically significant on the DbVPR gene expression being all the other factors, main and interactions, characterized by p-values higher than 0.05 (Table 5). In agreement with the RSM approach, DbVPR expression was maximizes (1.80-fold change) at 0.25 mg/L SO₂, pH 4.5, and 12.5% (v/v) ethanol, as observed for the DbCD.

The release of VPs was determined in the experimental conditions adopted in the Box–Behnken experimental design. Although 10 mg/L of each hydroxycinnamic acid were added to the SWM, the initial concentrations of p-coumaric acid and ferulic acid were estimated at 8.40 ± 0.07 mg/L and 6.71 ± 0.25 mg/L, respectively. As expected, these compounds proportionally decreased as the VPs increased (data not shown). The highest concentration of VPs was reached under a condition that is more permissive the yeast growth [SO₂ 0.125 mg/L, pH 3.5 and ethanol 5% (v/v)] in comparison to the expression of DbCD and DbVPR genes [SO₂ 0.25 mg/L, pH 4.5, and ethanol 12.5% (v/v)]. Indeed, VPs are released at a final concentration of 6.45 and 5.18 mg/L of ethyl phenol and ethyl guaiacol, respectively, in run 9 whereas DbCD and DbVPR genes were approximatively half of the expression values detected in run 4.

In general, some considerations arose from the calculated yields of VPs (Table 1). First, the lowest conversion of acids in the corresponding vinyl compounds was detected for the vinyl guaiacol that was mostly produced at trace level in all the analyzed runs (Table 1). We could speculate that this behavior could be linked to a higher activity of DbCDp toward the coumaric acid rather than the ferulic acid. On the contrary, ethyl phenol and ethyl guaiacol yields were found relatively balanced each other suggesting a similar capability of the DbVPR enzyme to transform its two substrates, the vinyl derivates. However, for this observation studies are required to analyze the activity of DbCDp and DbVPRp in the metabolic pathway of VPs under enological conditions.

Data processing by the RSM approach released four second-order equations with R-squared values indicating that the model as fitted explained 94.5, 81.2, 71.8, and 78.2% of the variability in vinyl phenol, vinyl guaiacol, ethyl phenol, and ethyl guaiacol molar ratios, respectively (calculated against the corresponding substrates of hydroxycinnamic acids). Main and interaction effects (linear and quadratic) of the factors on the VP production are reported in Table 5 and shown in the standardized Pareto charts (Figure 1). Considering the influence of individual factors, ethanol, and pH produced a significant effect (P-value < 0.05) on the production of such aromatic compounds whereas SO₂ did not result involved in. In particular, ethanol influenced the release of vinyl phenol, ethyl phenol, and vinyl guaiacol while pH was important in determining the variability of ethyl guaiacol. No linear interaction between factors resulted statistically significant for the synthesis of VPs.